RNA is not a “plain” molecule—it is replete with various chemical modifications. Just how many types of modifications exist, where they are located, and what they do is becoming clearer, as researchers hone the tools to examine such questions at the transcriptome-wide level.
To find new ideas, scientists sometimes go back in time—to studies published decades ago. That is what Wendy Gilbert did for a recent study on RNA modifications. Transfer RNAs (tRNAs) and messenger RNAs (mRNAs) teem with them—over 150 different variations of adenosine, cytosine, guanosine, and uridine. Pseudouridine is abundant in rRNA and tRNA, but when Gilbert began her study in 2013, nobody knew if the modification was present in mRNA.
Gilbert, an associate professor at Yale University, turned to a study published in 1993 by a pair of biochemists at Roche Research Center in New Jersey. Andrey Bakin and James Ofengand had added a bulky chemical group to pseudouridine in RNA. And they had shown that this supersized pseudouridine impeded the action of reverse transcriptase, the enzyme that generates complementary DNA (cDNA) from an RNA template, stopping the enzyme in its tracks.
Gilbert and her colleagues then applied the method to yeast and human cells, using next-generation sequencing (NGS) to interrogate the resulting cDNAs. The prematurely truncated cDNAs yielded a signature for pseudoruridine across the genome, detected on the transcripts of several hundred human genes. Several other groups emerged with mRNA maps of pseudouridine at about the same time. And the modification was not just a fluke—its levels on mRNA shifted in response to cellular stress, suggesting a regulatory role.
Studies like Gilbert’s are part of a wave of publications examining modifications at the genome-wide level. Several modifications other than those involving pseudouridine are under scrutiny, in particular N6-methyladenosine (m6A), which has a key role in regulating mRNA during development and disease. The methods being applied to mRNA and long noncoding RNAs are new, and many are still being tested in the crucible of scientific opinion and investigated through trial-and-error experimentation.
The new methods are also advancing the emerging field of epitranscriptomics, a loosely defined term that encompasses the biology of RNA modifications and their associated regulatory factors. “In the future, you won’t think of any RNA molecule as just an information-carrying molecule with some secondary structure,” says Christopher Mason, an associate professor at Weill Cornell Medicine in New York City. “You’ll think of it in terms of how it is modified, how it is regulated, how it is packaged, and how it is controlled.”
Moving to mRNA modifications
Research on the detection of tRNA and ribosomal (rRNA) modifications is a more mature field than that dealing with mRNAs. After all, these noncoding RNAs are more abundant than mRNA, facilitating biochemical studies. And tRNA in particular is amenable to mass spectrometry analyses, being small enough for researchers to piece together the chemistry of a modification and its location within the sequence. Moreover, many researchers view mRNA as a more daunting challenge, with its shifting cellular landscape of long and sometimes rare molecules.
To date, only a handful of modifications have been studied on mRNA, and some are highly contested—researchers argue heatedly about their prevalence, their biological relevance, and whether analyses to “call” a modification site are stringent enough.
Some of the open questions in the field will be resolved as existing techniques are honed and new ones invented. Among the methods used to find new modifications and probe existing ones, some of the most well-developed deploy antibodies to pull down RNAs containing modifications.
Writers, readers, and erasers
Research done in 2012 that mapped m6A using antibody pull-down methods is widely credited with opening up the field. Prior to 2012, there was evidence that m6A might have a key role in mRNA regulation. It had been found in a variety of eukaryotes, for instance at elevated levels in the mRNA of yeast undergoing sporulation, and in dividing tissues in Arabidopsis.
With the emergence of NGS methods, two research groups took a closer look, one led by Gideon Rechavi at the Sheba Medical Center, affiliated with Tel Aviv University in Israel, and the other by a team including Mason and Samie Jaffrey, a professor of pharmacology at Weill Cornell Medical College and a cofounder of Gotham Therapeutics, a New York-based biotech startup.
Both groups used m6A antibodies to pull down RNA fragments that contained the modification. They then created cDNA libraries from the fragments and sequenced them, generating a map of potential m6A sites. Their methods, dubbed meRIP-Seq and m6A-seq, uncovered thousands of modifications. Rechavi’s group, for instance, found more than 12,000 m6A sites in the transcripts of more than 7,000 human genes. m6A was also present on mammalian long noncoding RNAs.
Scientists have gone on to identify dedicated enzymes that interact with m6A: “writers” that add the modification to mRNA, “readers” that carry out its functions, and “erasers” that remove it. And they have shown that m6A can regulate mRNA translation, stability, and splicing, often in response to a stimulus. In Arabidopsis, m6A stabilizes transcripts involved in the salt and osmotic stress response. In mammals, it seems to have a role in learning and memory, in the generation of blood cells, and in X-chromosome inactivation, among other functions.
Findings such as these reflect a time in evolutionary history when RNA may have reigned as a central molecule, in what some biologists call the “RNA world.” Says Rechavi. “I am convinced that it will turn out that this is a major level of regulation of gene expression.”
“A druggable area”
These discoveries also have the potential for real-world impact. “We can see the first fruits of practical implications,” says Rechavi, who notes that several studies have linked the m6A modification to cancer. For instance, Tony Kouzarides and his colleagues at the Gurdon Institute in Cambridge, United Kingdom, recently showed that an m6A writer, METTL3 (methyltransferase-like 3), prompts the growth of acute myeloid leukemia cells. And though most are reticent about their exact focus, several biotech startups are moving into this area.
“This is a druggable area and there are good targets,” says Kouzarides, who is also a cofounder of Storm Therapeutics, a biotech company in Cambridge, UK, developing products based on findings from Kouzarides’ lab. Richard Gregory, founder of the RNA-focused company Twentyeight-Seven Therapeutics, in Cambridge, Massachusetts, explains that potential targets such as METTL3 are structurally similar to DNA methylases that are the targets of existing cancer drugs. Accent Therapeutics, based in Lexington, Massachusetts, is pursuing several RNA-modifying proteins, according to scientific cofounder Chuan He.
Meanwhile, researchers have adapted meRIP-Seq/m6A-seq to detect other modifications, for instance N1-methyladenosine (m1A), and more recently, N4-acetylcytidine (ac4C). And they are generating the next generation of antibody-based techniques, such as miCLIP, developed by Jaffrey and colleagues to detect m6A. The method involves crosslinking an antibody to an RNA modification, resulting in a block to reverse transcriptase and a truncated cDNA. The approach pinpoints the location of a modification more precisely than the 100–200 base-pair reads generated from meRIP-Seq/m6A-seq.
Rechavi cautions that such methods work best with antibodies validated to stick well to their intended target, with minimal off-target binding. But that is only one issue in a field where new methods and data are constantly being challenged.
Over the last few years, several whole-transcriptome maps for m6A and pseudouridine have been generated, and studies are emerging that assess the quality of the combined datasets as they become more reliable. But the data on some more-recently studied modifications—such as m1A and 2΄-O-methyl (Nm)—are still hotly debated.
Most methods in the field are “appropriate to detect some common ground at the lower end in terms of numbers, and then the opinions diverge,” says Mark Helm, an associate professor of pharmaceutical chemistry at the University of Mainz, Germany.
Controversy can arise when “interpretation of what is a signal versus not a signal is very different,” adds Michaela Frye, a professor at the German Cancer Research Center in Heidelberg and a consultant for Storm Therapeutics.
Several reviews have addressed such controversies, including pieces coauthored by Helm, Jaffrey, and others. These reviews outline approaches for setting up experimental controls, using orthogonal approaches to verify results, and ensuring more stringent data analysis. Researchers also call for more experiments to validate an RNA modification site as a true site with a biological function—for instance, by knocking it out or using a technique called SCARLET, developed by Tao Pan’s group at the University of Chicago in Illinois. SCARLET can biochemically detect and quantify a modified residue at any nucleotide in a specific transcript. However, such validation is “almost never done,” says Gilbert.
But despite the back-and-forth that is common to any emerging field, there is one area of consensus—the need for new experimental approaches.
Old becomes new
He’s goal is to generate new epitranscriptomic methods that are both quantitative and accessible. To meet this challenge, He, a professor of chemistry at the University of Chicago, has a five-year, USD 10.6-million grant from the National Human Genome Research Institute. “The major need for the field is new technology,” says He.
Researchers ultimately wish for a method that could show precisely where on the genome a modification sits and also quantify what percentage of mRNAs and noncoding RNAs have that modification. Methods with such potential include emerging techniques to sequence individual RNAs, such as nanopore sequencing—which involves using an electric current to thread an RNA molecule through a tiny pore in a membrane, then reading the sequence by measuring changes in the current.
A new study in BioRxiv, by a team including Mason and Eva Maria Novoa of the Center for Genomic Regulation in Barcelona, Spain, intriguingly shows 90% accuracy at detecting m6A with nanopore sequencing.
Some researchers are optimistic that accuracy will improve substantially with the nanopore technique, but others are hesitant. “I would love the method to work because I think it would be the solution to everything,” says Frye, but she notes that detecting extremely rare modifications is a challenge, even with a method that is only slightly inaccurate.
Other longer-shot approaches include improving mass spectrometry methods that could potentially be applied to mRNA—or imaging techniques, which are still at nascent stages— says Sara Rouhanifard, an assistant professor of bioengineering at Northeastern University in Boston.
He’s group is currently working on techniques they can use in tandem with next-generation or other sequencing methods. “Hopefully in about two years, people can just buy a kit from a company” for a variety of methods, says He, “and use software they can download, and do everything themselves.” His group, for instance is applying evolutionary selection to generate reverse transcriptases that introduce a base change into cDNA to detect different RNA modifications.
Meanwhile, Gilbert is still going back in time. She is not alone, she says, in looking through old studies to see what kind of methods she can use to “turn into a next-generation sequencing assay [that will help] find a new modification.” She also is eager to explore how RNA-binding proteins modulate and interpret information conveyed by pseudouridine, which does not seem to have a dedicated mRNA reader or writer, she adds.
Current and future tools
More researchers outside the field have begun to show interest in epitranscriptomic techniques, particularly for the m6A modification.
Several kits are available for meRIP-Seq and m6A-seq (e.g., from New England Biolabs). Jaffrey uses antibodies from Abcam and Synaptic Systems. Other techniques such as miCLIP require a bit more finesse, adds Jaffrey, who has published detailed protocols on miCLIP and meRIP-Seq.
Schraga Schwartz, an RNA researcher at the Department of Molecular Genetics at the Weizmann Institute of Science, in Rehovot, Israel, notes that going deep into the data requires expertise in computational biology, though he believes that as the tools evolve, the data will become easier to interpret. Researchers interested in m6A can “kind of get a general sense of what the data look like,” he says, but generating quantitative data across different conditions “gets a lot trickier.”
Mason encourages researchers to expand their skills into epitranscriptomics. “Rarely in biology do you have such a wide-open field that is ripe for discovery. In this case the question is, ‘Does modification of my RNA impact some part of my biology?’ The answer so far seems to almost always be ‘yes,’” he says. “I think the advantage to picking up this skill set is that you can quickly bring a new ray of light that could illuminate some part of a biological question you have had for a long time, and then see it in a new way.”
Newly offered instrumentation, apparatus, and laboratory materials of interest to researchers in all disciplines in academic, industrial, and governmental organizations are featured in this space. Emphasis is given to purpose, chief characteristics, and availability of products and materials. Endorsement by Science or AAAS of any products or materials mentioned is not implied. Additional information may be obtained from the manufacturer or supplier.